HPGPT: Conversational Multi-Agentic AI Chatbot

📌 Overview

HPGPT is a conversational AI platform that provides multi-agent, domain-specific assistance across coding, analytics, database querying, document understanding, and web search—all through a single chat interface.

The system leverages LangChain, LangGraph, Google Gemini (via ADK), and Groq LLaMA models to orchestrate specialized agents:

• 🤖 General Assistant – fallback & casual queries
• 💻 Coding Agent – generates executable code (Python, C, C++, Java, JS, HTML)
• 📊 Analytics Agent – processes CSV/XLSX, runs pandas queries, renders Plotly charts
• 🗄️ Database Agent – executes SQL queries against a linked PostgreSQL/MySQL database and returns structured results or natural language summaries
• 📑 Document Agent – RAG-powered document Q&A (PDF, Word, Excel, TXT)
• 🌐 Websearch Agent – real-time factual search via Tavily API

---

⚡ Features

• Multi-Agent Orchestration with LangGraph
• Database Querying Agent – write & execute SQL securely against a live DB
• Context persistence with PostgreSQL + LangChain Memory
• File-aware Q&A (PDF, Excel, CSV, DOCX, TXT)
• Real-time WebSocket chat with streaming responses
• Interactive Plotly visualizations in chat
• Secure file handling with session-based storage
• Scalable & extensible agent pipeline

---

🏗️ System Architecture

• Frontend: Flask + JS (chat UI, file uploads, streaming charts, syntax highlighting)
• Backend: FastAPI (agent routing, WebSocket streaming, file processing, database queries)
• LangGraph: session/context manager + agent dispatcher
• Database: PostgreSQL (users, sessions, messages, feedback, SQL query execution)
• Agents: Modular Python agents powered by Gemini/Groq

---

📂 Project Structure

HPGPT/
│── backend/
│   ├── agents/ 
│   │   ├── analytics_agent.py
│   │   ├── coding_agent.py
│   │   ├── document_agent.py
│   │   ├── websearch_agent.py
│   │   └── database_agent.py
│   │
│   ├── utils/ 
│   │   ├── groq_client.py
│   │   ├── langgraph_manager.py
│   │   ├── langgraph_pipeline.py
│   │   ├── file_processor.py
│   │   └── file_utils.py
│   │
│   ├── database/
│   │   ├── db_manager.py   # connection & query execution
│   │   └── auth.py         # authentication & sessions
│   │
│   └── main.py  # FastAPI entrypoint
│
│── frontend/
│   ├── app.py  # Flask server
│   ├── templates/index.html
│   ├── static/js/main.js
│   ├── static/css/styles.css
│
│── requirements.txt
│── README.md

---

🚀 Installation & Setup

0. Prerequisites

• Python ≥ 3.10
• PostgreSQL/MySQL running locally or remote
• VS Code / IDE recommended

1. Clone Repository

git clone https://github.com/CharithKalasi/HPGPT.git
cd HPGPT

2. Create Virtual Environment

python -m venv venv

# Linux / Mac
source venv/bin/activate   

# Windows PowerShell
.\venv\Scripts\Activate
# (If activation fails, run this first to allow script execution)
Set-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy RemoteSigned

3. Install Dependencies

pip install -r requirements.txt

4. Configure Database

Update .env with PostgreSQL/MySQL credentials. The system will auto-connect and manage tables via db_manager.py.

5. Run Backend (FastAPI)

cd backend
uvicorn main:app --reload

6. Run Frontend (Flask)

cd frontend
python app.py

7. Open in Browser

http://127.0.0.1:5000/

---

💡 Usage Examples

🤖 General Assistant

Prompt:

What's the capital of France?

Response:

The capital of France is Paris.

💻 Coding Agent

Prompt:

Write a Python function to check if a number is prime.

Response:

def is_prime(n):
    if n <= 1:
        return False
    for i in range(2, int(n**0.5) + 1):
        if n % i == 0:
            return False
    return True

📊 Analytics Agent

Prompt:

Upload sales.csv and show me the total revenue by product category in a bar chart.

Response:
Interactive Plotly bar chart with revenue grouped by category.

🗄️ Database Agent

Prompt:

Show me the top 5 customers by purchase amount.

Generated SQL:

SELECT customer_name, SUM(amount) AS total_spent
FROM orders
GROUP BY customer_name
ORDER BY total_spent DESC
LIMIT 5;

Response:

Customer Name	Total Spent
Alice	15,200
Bob	12,450
Charlie	9,880

📑 Document Agent

Prompt:

Summarize the attached PDF in 5 bullet points.

Response:

• Extracted key points from PDF...

🌐 Websearch Agent

Prompt:

What's the latest news about electric vehicles in India?

Response:
Latest web snippets summarizing EV adoption and government policies.

---

📊 Results

• Unified conversational interface for multi-domain tasks
• Automatic agent routing without dropdown/manual selection
• Reliable RAG-based document Q&A
• Real-time code execution & data visualization
• Secure SQL query execution via Database Agent

---

✅ Conclusion

HPGPT bridges the gap between general-purpose chatbots and enterprise-level intelligent assistants.
Its modular, multi-agent architecture ensures that each agent—whether for documents, analytics, coding, research, or more—works in unison to handle complex tasks through simple natural language prompts.

This design makes HPGPT both scalable and adaptable, empowering users across domains to unlock actionable intelligence without technical barriers.

---

🔗 References & Resources

• 📂 Code: GitHub Repo
• 🎥 Demo: Sample Video
• 📚 Docs:
  • LangChain
  • LangGraph
  • Flask
  • Google AI SDK
  • ChromaDB